Euler and the Ongoing Search for Odd Perfect Numbers

Brian D. Beasley
Presbyterian College, Clinton, SC

Abstract

Leonhard Euler, after proving that every even perfect number has the form given by Euclid, turned
his attention to finding odd perfect numbers. Euler established a basic factorization pattern that
every odd perfect number must have, and mathematicians have expanded upon this Eulerian form
ever since. This talk will present a brief summary of Euler’s result and some recent generalizations.
It will also note connections between odd perfect numbers and the abundancy index (the abundancy
index of a positive integer is the ratio of the sum of its positive divisors to itself). In particular,
finding a positive integer with an abundancy index of 5/3 would finally produce that elusive odd
perfect number.

1 Before Euler
“As for me, I judge that one can find real odd perfect numbers.”

– René Descartes

1.1 Euclid and Subsequent Conjectures

As noted in a number of historical accounts (including [1], [2], and [4]), the study of perfect
numbers dates back even before the time of Euclid (ca. 300 B.C.). Euclid provided the
following definition: A perfect number is equal to its parts. That is, a number is perfect
if it equals the sum of its proper (aliquot) positive divisors. In addition, the first recorded
result concerning perfect numbers is due to Euclid [5] (Book IX, Proposition 36).

Theorem 1. If 2n − 1 is prime, then 2n−1 (2n − 1) is perfect.

Using n = 2, 3, 5, and 7 in Euclid’s formula produces the first four perfect numbers, well
known to the ancients:

6, 28, 496, 8128, ...

Based on this short list, two conjectures about perfect numbers proved irresistible. Nico-
machus (ca. 60-120) believed that perfect numbers alternate ending in 6 and 8, while
Iamblichus (ca. 283-330) claimed that for each positive integer k, there is exactly one per-
fect number with k decimal digits [2]. Alas, both conjectures are false, as the next two
perfect numbers are 33,550,336 and 8,589,869,056. On the other hand, a more intriguing
question, also from Nicomachus [2], remains open to this day.

ACMS 19th Biennial Conference Proceedings, Bethel University, 2013 21

Conjecture 1. Euclid’s rule gives all perfect numbers; in particular, no odd number is
perfect.

In addition to his conjecture about the nonexistence of odd perfect numbers, Nicomachus,
along with Theon of Smyrna (ca. 70-135), distinguished between deficient and abundant
numbers [2]. In modern notation, given a positive integer n, we let s(n) be the sum of
the proper positive divisors of n. Then Euclid’s definition translates into: n is perfect
if s(n) = n. Similarly, n is deficient if s(n) < n, while n is abundant if s(n) > n. In
particular, we note that every prime power is deficient, since 1 + p + p2 + · · · + pk−1 < pk for
any prime p and any positive integer k; hence there are infinitely many deficient numbers.
The smallest abundant numbers are 12 and 18, which happen to be nontrivial multiples of
the first perfect number, although the next largest abundant number (20) is not a multiple
of any perfect number.

Motivated by such examples, and the fact that every abundant number smaller than 900
is even, Jordanus Nemorarius (1225-1260) made the following claims [2].

Conjecture 2. (a) Every (nontrivial) multiple of a perfect number is abundant.

(b) Every (nontrivial) divisor of a perfect number is deficient.
(c) No odd number is abundant.

In fact, the first two of these conjectures are true, but the third is false, since 945 is
abundant. Since every multiple of an abundant number is also abundant, we may conclude
that there are infinitely many abundant numbers, including infinitely many odd ones. This
observation at least provides us with hope that there are odd perfect numbers, but we still
have no proof or disproof of the existence of such a number.

1.2 Setting the Stage for Euler: Descartes and Frenicle

Interest in studying perfect numbers increased in the 1600s, with a particular focus on dis-
proving the conjecture of Nicomachus. For example, René Descartes (1596-1650) believed
that he could prove the following claims [2].

Conjecture 3. (a) Euclid’s rule gives all even perfect numbers.

(b) Every odd perfect number has the form ps2 with p prime.

We continue to follow Dickson’s account in [2], examining the exchange between Descartes
and one of his frequent correspondents, Bernard Frenicle de Bessy (1605-1675). Frenicle
agreed with Descartes’ conjecture about odd perfect numbers and further claimed that
this prime p must be congruent to 1 mod 4. However, neither Descartes nor Frenicle
offered a proof of either claim. To demonstrate his conviction that an odd perfect number

ACMS 19th Biennial Conference Proceedings, Bethel University, 2013 22

exists, Descartes noted that if p = 22, 021 were prime, then taking s = 3 · 7 · 11 · 13 would
produce the odd perfect number ps2 ; alas, as Descartes knew, we have the factorization
22, 021 = 192 · 61. Undaunted, Descartes wrote to Frenicle to suggest that replacing 7 or
11 or 13 in s might eventually work instead. Indeed, both Descartes and Frenicle found
several examples of even multiperfect numbers – but no odd perfect numbers ...

2 Euler
“Whether ... there are any odd perfect numbers is a most difficult question.”

– Leonhard Euler

2.1 Completing the Work of Euclid on Even Perfect Numbers

As Dunham notes in [4], a letter from Christian Goldbach in 1729 may have initiated
Leonhard Euler’s work in the field of number theory, inspiring Euler to tackle the following
claim by Pierre de Fermat.
n
Conjecture 4. For each nonnegative integer n, Fn = 22 + 1 is prime.

This sequence of so-called Fermat numbers begins 3, 5, 17, 257, 65537, etc. In particular,
Fn is indeed prime for 0 ≤ n ≤ 4. But Euler showed the claim is false when n = 5, as 641
divides F5 = 4, 294, 967, 297. In fact, the reality about Fermat numbers could ultimately
prove to be the exact opposite of the original conjecture, as no other primes have been
found among Fn for n ≥ 5 [1].

It is indeed an understatement to say that Euler went on to make many contributions

to number theory. One example was the search for amicable numbers, a pair of positive
integers m and n with s(m) = n and s(n) = m. Only three pairs of amicable numbers were
known before Euler, yet he proceeded to discover 59 new pairs [4]. Another example was
Euler’s introduction of the function φ to enumerate the positive integers not exceeding a
given natural number n which are relatively prime to n. Euler applied this new function
to generalize Fermat’s Little Theorem, which states that if p is prime and gcd(a, p) = 1,
then ap−1 ≡ 1 (mod p). Following is Euler’s generalization of Fermat’s result [1].

Theorem 2. Given the positive integer n, if a is any integer with gcd(a, n) = 1, then
aφ(n) ≡ 1 (mod n).

In keeping with his usual creativity in applying new approaches to old problems, Euler
reached a breakthrough in studying perfect numbers, a seemingly simple observation that
nevertheless produced profound results. Instead of using s(n) to sum the proper positive

ACMS 19th Biennial Conference Proceedings, Bethel University, 2013 23

 R
divisors of n, Euler introduced the notation n to represent the sum of all positive
R divisors
of n, including n itself. The modern notation for this function uses σ to replace , so that
we now write σ(n) = s(n) + n.

Euler was able to show that σ is a multiplicative number theoretic function [1]; that is,
if gcd(m, n) = 1, then σ(mn) = σ(m)σ(n). This property proved quite valuable for both
computational and analytical use. For example, we may calculate σ(360) = σ(23 · 32 · 5) as
follows:

σ(360) = σ(23 )σ(32 )σ(5)

= (1 + 2 + 22 + 23 )(1 + 3 + 32 )(1 + 5)
= (15)(13)(6)
= 1170.

We note that if the product of the three factors in the second line is expanded via distri-
bution, then the resulting 24 terms are precisely the positive divisors of 360.

More importantly, Euler used the multiplicative property of σ to prove the long-awaited
converse of Euclid’s theorem on even perfect numbers.

Theorem 3. Euclid’s rule gives all even perfect numbers.

In particular, Euler made use of the ratio σ(n)/n, which we now call the abundancy index
I(n) of n. We will return to the abundancy index later, but for now, we simply note that
n is perfect if and only if I(n) = 2; similarly, n is deficient when I(n) < 2 and is abundant
when I(n) > 2.

Proof. We follow Euler’s proof as outlined in [1] and [4]. If N is an even perfect number,
then write N = 2k−1 b with b odd and k > 1. Then 2N = σ(N ) = 2k b, while σ(N ) =
σ(2k−1 )σ(b) = (2k − 1)σ(b). Thus
σ(b) 2k
= k .
b 2 −1
Since 2k − 1 and 2k are relatively prime, we know that 2k − 1 divides b. (At first, Euler
seems to have believed that this observation was enough to conclude b = 2k − 1, but he
later corrected the gap in his proof as follows; see [2] and [16].) Then b = (2k − 1)a for
some positive integer a, which implies σ(b) = 2k a. But both a and b are positive divisors
of b, so 2k a = σ(b) ≥ a + b = 2k a and thus σ(b) = a + b. This yields a = 1, so σ(b) = 1 + b
and hence b is prime.

ACMS 19th Biennial Conference Proceedings, Bethel University, 2013 24

2.2 The Eulerian Form of an Odd Perfect Number
In addition to establishing that Euclid’s rule produces all even perfect numbers, Euler was
able to prove that every odd perfect number must have the following form [1].

Theorem 4. If an odd perfect number exists, then it has the form pk s2 , where p is prime,
gcd(p, s) = 1, and p ≡ k ≡ 1 (mod 4).

Before proceeding with Euler’s proof, we pause to note that his result was not quite what
Descartes and Frenicle had conjectured, as they believed that k = 1, but it came very
close. In fact, current research continues in an effort to prove k = 1. For example, Dris
has made progress in this direction, although his paper refers to Descartes’ and Frenicle’s
claim (that k = 1) as Sorli’s conjecture [3]; Dickson has documented Descartes’s conjecture
as occurring in a letter to Marin Mersenne in 1638, with Frenicle’s subsequent observation
occurring in 1657 [2].

Proof. We outline the proof of Euler given in [1], noting that once again, Euler employed
the σ function in his argument. If N is an odd perfect number, then write N = pk11 pk22 · · · pkr r
with each pi an odd prime and each ki > 0. Thus
2N = σ(pk11 )σ(pk22 ) · · · σ(pkr r ).
k
Since 2N ≡ 2 (mod 4), exactly one factor σ(pj j ) is congruent to 2 modulo 4, with the
other factors all being odd. Also, since σ(pki i ) = 1 + pi + p2i + · · · + pki i , we may consider
cases for pi modulo 4:
k
(a) If pi ≡ 1 (mod 4), then σ(pki i ) ≡ ki +1 (mod 4). Since σ(pj j ) ≡ 2 (mod 4), we conclude
kj ≡ 1 (mod 4). But every other σ(pki i ) is odd, so ki must be even for i 6= j.

(b) If pi ≡ 3 ≡ −1 (mod 4), then σ(pki i ) is 0 modulo 4 if ki is odd and is 1 modulo 4 if ki

is even, which immediately implies that pj ≡ 1 (mod 4). But 4 cannot divide σ(pki i ), so ki
must be even for any prime pi which is congruent to 3 modulo 4.

With this theorem concerning the necessary form of any odd perfect number, Euler pre-
pared the way for future mathematicians to refine his result and to continue progress toward
a proof or disproof of the existence of odd perfect numbers. Just over one hundred years
after Euler’s death, another mathematician would indeed contribute significantly to the
list of conditions needed for an odd perfect number.

ACMS 19th Biennial Conference Proceedings, Bethel University, 2013 25

3 After Euler
“ ... a prolonged meditation on the subject has satisfied me that the existence of any one
such [odd perfect number] – its escape, so to say, from the complex web of conditions
which hem it in on all sides – would be little short of a miracle.”

– James Joseph Sylvester

3.1 Sylvester’s Web

In 1888, James Joseph Sylvester picked up where Euler left off, using the Eulerian form
of an odd perfect number to establish a number of important results [6]. As Dickson [2]
noted, in that year alone, Sylvester was able to prove:

• No odd perfect number is divisible by 105.

• An odd perfect number must have at least four distinct prime divisors. (Sylvester
proved later that year that an odd perfect number must have at least five distinct
prime divisors, and he conjectured that at least six were required.)

• If an odd perfect number is not divisible by 3, then it must have at least eight distinct
prime divisors.

In addition to establishing this “web” of requirements for odd perfect numbers, Sylvester
emphasized the importance of resolving such a question that dated back to ancient times.
He referred to the issue as being a “problem of the ages comparable in difficulty to that
which previously to the labors of Hermite and Lindemann environed the subject of the
quadrature of the circle” [2].

Inspired by Sylvester’s work, mathematicians have endeavored ever since to extend the
web of conditions for odd perfect numbers. We outline a small sample of such conditions.
For example, it took 37 years after Sylvester’s claim before Gradstein proved that an
odd perfect number must have at least six distinct prime divisors; that lower bound was
subsequently improved to seven by Robbins and Pomerance (independently) in 1972 and
to eight by Hagis in 1980 [19]. (Chein also proved the result for eight in 1979, but his
dissertation was not published [6].) The current best known lower bound is nine, due to
Nielsen in 2007 [12]. Nielsen also established that if an odd perfect number is not divisible
by 3, then it has at least twelve distinct prime divisors [12].

Next, we follow the progress listed in [9] on finding lower bounds on the largest prime
factor of an odd perfect number:
60 – Kanold, 1944

ACMS 19th Biennial Conference Proceedings, Bethel University, 2013 26

 11,200 – Hagis and McDaniel, 1973
105 – Hagis and McDaniel, 1975
3 × 105 – Condict, 1978
5 × 105 – Brandstein, 1982
106 – Cohen and Hagis, 1998
107 – Jenkins, 2003

Since the publication of [9], the lower bound on the largest prime divisor of an odd perfect
number has been improved, to 108 by Goto and Ohno in 2008 [7]. Iannucci has also
shown that the second largest prime factor of an odd perfect number must be at least 104 ,
improving on the previous results of 139 (Pomerance, 1975) and 103 (Hagis, 1981) [9]. In
addition, Iannucci has proved that the third largest prime factor of an odd perfect number
must be at least 100 [10].

We also note some of the congruence conditions that apply to odd perfect numbers. We have
already encountered Euler’s result that every odd perfect number must be congruent to 1
modulo 4. In 1953, Touchard established that an odd perfect number must be congruent
to either 1 modulo 12 or 9 modulo 36 [18]. In 2008, Roberts refined this result, proving
that every odd perfect number must be congruent to either 1 modulo 12, 117 modulo 468,
or 81 modulo 324 [14].

Since Sylvester’s time, mathematicians have woven more and more strands in the “web”
which surrounds odd perfect numbers. Indeed, it would be quite difficult to list all of the
additional conditions now known. Instead, we conclude this section by simply noting the
recent contributions of Ochem and Rao: In 2012, they showed that an odd perfect number
must be greater than 101500 , must be the product of at least 101 (not necessarily distinct)
prime factors, and must have a prime power divisor greater than 1062 [13].

3.2 Connections with the Abundancy Index

We recall that Euler applied the ratio σ(n)/n in his proof that Euclid’s rule gives all
even perfect numbers. Using the modern notation and terminology I(n) = σ(n)/n for
the abundancy index of a positive integer n, we summarize in this section some of the
connections between abundancy results and the search for odd perfect numbers.

To generalize an earlier definition, given an integer k ≥ 2, we call the positive integer

n multiperfect or k-perfect if I(n) = k. Descartes [2] established a number of results
concerning multiperfect numbers, including the following connection between 3-perfect
and 4-perfect numbers.

ACMS 19th Biennial Conference Proceedings, Bethel University, 2013 27

Theorem 5. If n is 3-perfect and 3 does not divide n, then 3n is 4-perfect.

Proof. We observe that since σ is a multiplicative function, so is I. Consequently, since 3

does not divide n, we have
4
I(3n) = I(3)I(n) = 3 · 3 = 4.

In 2000, Weiner [20] showed a remarkable connection between a specific abundancy index
and the existence of an odd perfect number. We give his theorem and proof here, noting
the similarity to the method used by Descartes in the previous result.

Theorem 6. If there is a positive integer n with I(n) = 5/3, then 5n is an odd perfect
number.

Proof. We outline the key steps in Weiner’s proof [20]. First, since 3σ(n) = 5n, 3 must
divide n. But then 2 cannot divide n, as otherwise, 6 would also, yielding the contradiction
that n must be perfect or abundant. Hence n is odd, which implies that σ(n) is also odd.
In particular, this means that each prime factor of n must occur to an even power [1],
so 32 must divide n. But then 5 cannot divide n, as otherwise, 45 would also, yielding
the contradiction
P that I(n) ≥ I(45) = 26/15 > 5/3. (Here, we have used the fact that
I(n) = d|n (1/d) (for example, see [1]), which in turn implies that if d is a positive divisor
of n, then I(d) ≤ I(n).) Thus
6 5
I(5n) = I(5)I(n) = 5 · 3 = 2.

In 2003, Ryan [15] provided the following generalization of Weiner’s theorem.

Theorem 7. If there exist positive integers m and n such that m is odd, 2m − 1 is prime,
2m−1 does not divide n, and I(n) = (2m−1)/m, then n(2m−1) is an odd perfect number.

Ryan also showed in [15] that if m is even but not a power of 2, then I(n) = (2m − 1)/m
has no solution. Three years later, Holdener [8] proved another generalization of both
Weiner’s and Ryan’s theorems, giving the following necessary and sufficient condition for
the existence of an odd perfect number.

Theorem 8. There is an odd perfect number if and only if there are positive integers p,
n, and k such that p is prime, p does not divide n, p ≡ k ≡ 1 (mod 4), and

2pk (p − 1)
I(n) = .
pk+1 − 1

ACMS 19th Biennial Conference Proceedings, Bethel University, 2013 28

We note the direct connection between Holdener’s result and the Eulerian form of an odd
perfect number: If such positive integers p, n, and k do exist, then pk n is an odd perfect
number, since
pk+1 − 1 2pk (p − 1)
I(pk n) = I(pk )I(n) = · = 2.
pk (p − 1) pk+1 − 1
In particular, n would equal s2 in Euler’s result.

In the opposite direction, since the non-existence of such a value of I(n) would prove
that no odd perfect numbers exist, Holdener and Stanton [17] have studied these rational
numbers outside the range of I and have given them a descriptive name: Given a rational
number r > 1, r is an abundancy outlaw if I(n) 6= r for every positive integer n. Applying
Holdener’s theorem, we note for example that if 5/3, 13/7, or 17/9 is not an abundancy
outlaw, then an odd perfect number exists. The “outlaw” status of these three rational
numbers, along with others of the form in Holdener’s result, remains unknown [17].

Such observations raise the natural question of the distribution of abundancy indices vs.
the distribution of abundancy outlaws in [1, ∞). On the one hand, Laatsch [11] proved in
1986 that {I(n) : n ∈ N} is dense in [1, ∞), which raises the hope that perhaps 5/3 is in
fact an abundancy index. On the other hand, Weiner [20] showed in 2000 that the set of
abundancy outlaws is also dense in [1, ∞).

While the question of the existence of odd perfect numbers remains open, we may apply
Laatsch’s result about the density of abundancy indices to create some interesting examples.
In particular, we can make I(n) arbitrarily close to certain “famous” numbers:

• I(3 · 17 · 577 · 665857) = 1.414213562371 . . .

• I(2 · 3 · 5 · 11 · 29 · 277 · 67927 · 204109349) = 2.71828182845903 . . .
• I(2 · 3 · 5 · 7 · 11 · 23 · 163 · 23509 · 690120229) = 3.141592653589792 . . .

In addition, we are able to make I(n) arbitrarily close to 2 for odd n:

• I(3 · 5 · 7 · 11) = 1.9948 . . .

• I(3 · 5 · 7 · 11 · 389) = 1.99993 . . .
• I(3 · 5 · 7 · 11 · 383) = 2.00001 . . .
• I(3 · 5 · 7 · 11 · 389 · 29959) = 1.99999998 . . .

But the main question still remains, handed down to us by Euclid, Descartes, Euler,
Sylvester, and many others: Can we make I(n) equal to 2 for odd n???

ACMS 19th Biennial Conference Proceedings, Bethel University, 2013 29

4 Bibliography
4.1 References

1. D. Burton, Elementary Number Theory (7th Edition), McGraw-Hill, 2011.

2. L. Dickson, History of the Theory of Numbers, Volume 1: Divisibility and Primality,

Chelsea Publishing Co., 1952.

3. J. Dris, The Abundancy Index of Divisors of Odd Perfect Numbers, Journal of Integer
Sequences, Vol. 15 (2012), Article 12.4.4.

4. W. Dunham, Euler: The Master of Us All, MAA (The Dolciani Mathematical Ex-
positions, No. 22), 1999.

5. Euclid, The Thirteen Books of Euclid’s Elements (translated from Heiberg, with
introduction and commentary by Sir Thomas L. Heath, 2nd Edition), Dover, 1956.

6. S. Gimbel and J. Jaroma, Sylvester: Ushering in the Modern Era of Research on Odd
Perfect Numbers, Integers: Electronic Journal of Combinatorial Number Theory, Vol.
3 (2003), #A16.

7. T. Goto and Y. Ohno, Odd Perfect Numbers Have a Prime Factor Exceeding 108 ,
Mathematics of Computation, Vol. 77 (2008), 1859-1868.

8. J. Holdener, Conditions Equivalent to the Existence of Odd Perfect Numbers, Math-

ematics Magazine, Vol. 79, No. 5 (2006), 389-391.

9. D. Iannucci, The Second Largest Prime Divisor of an Odd Perfect Number Exceeds
Ten Thousand, Mathematics of Computation, Vol. 68, No. 228 (1999), 1749-1760.

10. D. Iannucci, The Third Largest Prime Divisor of an Odd Perfect Number Exceeds
One Hundred, Mathematics of Computation, Vol. 69, No. 230 (1999), 867-879.

11. R. Laatsch, Measuring the Abundancy of Integers, Mathematics Magazine, Vol. 59,
No. 2 (1986), 84-92.

12. P. Nielsen, Odd Perfect Numbers Have At Least Nine Distinct Prime Factors, Math-
ematics of Computation, Vol. 76 (2007), 2109-2126.

13. P. Ochem and M. Rao, Odd Perfect Numbers Are Greater Than 101500 , Mathematics
of Computation, Vol. 81 (2012), 1869-1877.

14. T. Roberts, On the Form of an Odd Perfect Number, Australian Mathematical Society
Gazette, Vol. 35, No. 4 (2008), 244.

ACMS 19th Biennial Conference Proceedings, Bethel University, 2013 30

 15. R. Ryan, A Simpler Dense Proof Regarding the Abundancy Index, Mathematics
Magazine, Vol. 76, No. 4 (2003), 299-301.

16. E. Sandifer, How Euler Did It: Odd Perfect Numbers, MAA Online, No. 37 (Nov.
2006), 1-5.
σ(N )+t
17. W. Stanton and J. Holdener, Abundancy “Outlaws” of the Form N , Journal of
Integer Sequences, Vol. 10, No. 9 (2007), Article 07.9.6.

18. J. Touchard, On Prime Numbers and Perfect Numbers, Scripta Mathematica, Vol.
19 (1953), 35-39.

19. S. Wagon, The Evidence: Odd Perfect Numbers, Mathematical Intelligencer, Vol. 7,
No. 2 (1985), 66-68.

20. P. Weiner, The Abundancy Ratio, a Measure of Perfection, Mathematics Magazine,

Vol. 73, No. 4 (2000), 307-310.

4.2 Helpful Web Sites

1. MacTutor History of Mathematics, J. J. O’Connor and E. F. Robertson:

http://www-history.mcs.st-and.ac.uk/HistTopics/Perfect numbers.html

2. OddPerfect.org:
http://www.oddperfect.org

3. Wolfram/MathWorld, C. Greathouse and E. Weisstein:

http://mathworld.wolfram.com/OddPerfectNumber.html

ACMS 19th Biennial Conference Proceedings, Bethel University, 2013 31